Beyond input-output mappings

In this post, a major paradigm shift in neuroscience from the last 20 years or so is described.  The move is from consideration of the activity of nervous systems as mapping from input to output (as in feed-forward neural networks) to a view whereby sensory input modulates the ongoing endogenous activity of the system (think of a recurrent network that is spontaneously generating activity before any input arrives) .  Sensory input is thus a perturbation to the system, whose effect will depend on the structure and activity of the system, rather than a stimulus producing a determinate response.


The article describes a specific example from a very simple animal, C. Elegans, which is a small worm that has taught us much about nervous systems.  Crucially, the simplicity of this animal makes it possible to identify paths linking sensory inputs and motor outputs.  Sensory input does affect motor output (the likelihood of the worm backing away from an odour source is reduced when an appetitive odour is sensed), but that effect is mediated and kept flexible through the action of intermediaries, the interneurons, which serve to ensure that the response is a flexible modulation of ongoing activity, rather than an invariant and passive "response".

The consequences of this shift in view for the interpretation of the central nervous system in other animals, ourselves included, are worthy of consideration.